Unable to set Codec to Custom Type (pguint)

* **asyncpg version**: 0.20.0
* **PostgreSQL version**: 12.2
* **Do you use a PostgreSQL SaaS? If so, which? Can you reproduce
the issue with a local PostgreSQL install?**: No, N/A, N/A
* **Python version**: 3.8.0
* **Platform**: Linux 64bit (Ubuntu 18.04)
* **Do you use pgbouncer?**: No
* **Did you install asyncpg with pip?**: Yes
* **If you built asyncpg locally, which version of Cython did you use?**: N/A
* **Can the issue be reproduced under both asyncio and
[uvloop](https://github.com/magicstack/uvloop)?**: That's actually what I'm using.

I am using the user extension [pguint](https://github.com/petere/pguint) (there's a pg11&12 branch) and it's installed fine. I'm trying to load ~100GB of article data onto my database, so I'm using the binary copy functions.

I created a binary encoder and decoder:
``` python
await con.set_type_codec( 'uint1', encoder=lambda x: struct.pack('!B',x), decoder=lambda x: struct.unpack('!B',x), format='binary')
```
But when I try to use it, I get:
```
Caught Insert Error. UID:2419 EX: no binary input function available for type uint1
```
as my error. I tried delving into the source of this project but wasn't able to glean much. Does this error mean that the postgres extension itself is missing the required function on its end?

As a side note: I wrote a correct encoder/decoder (both text and binary) for the `tsvector` type, I want to submit a pull request but have no idea where I should put the new code, want to point me in the right direction?

also, should I be including the postgres `varlena` header in my binary encoding function outputs? or does your code do that?

Start with set_type_codec and the binary COPY functions described in the report, reproducing the uint1 failure with PostgreSQL 12.2, asyncpg 0.20.0, Python 3.8, and uvloop. Check the pguint binary input behavior and the question about varlena headers. Done means the cause and supported guidance are established, including where a tsvector codec contribution would belong.

Written by the indexing model from the issue text.